互信息作為衡量兩幅圖像配準的相似性測度函數(shù)，當兩幅圖像配準時，互信息達到最大值。該文提出了基于互信息腦功能磁共振圖像配準新方法，采用了無需計算梯度的Powell直接搜索算法，并通過多分辨的方式加快了圖像配準速度。磁共振的配準實驗證明，互信息法能準確地實現(xiàn)多模態(tài)醫(yī)學圖像的配準，并且能達到亞像素的精度。
關 鍵 詞 功能磁共振; 圖像配準; 互信息; 亞像素

Abstract The functional magnetic resonance imaging (fMRI) image registration is a key technology in the study of the human brain function. In this article, we present a new fMRI image registration algorithm based on the mutual information. The mutual information is specified as scaling the similarity of two images registration. The mutual information of two images reaches to maximization when the two images are registered. Powell direct searching and multi-scale algorithm are introduced to accelerate the image registration speed without calculating gradient algorithm. Simulative experiments of fMRI image registration demonstrate that this method can precisely achieve the multimodal medical image registration, and can reach sub-pixel precision
Key words functional magnetic resonance imaging (fMRI); image registration; mutual information; sub- pixel